Poster Presentation 35th Lorne Cancer Conference 2023

Identifying Mesothelioma Prognostic Gene Signature Using Genetically Diverse CC-MexTAg Mouse Model (#113)

Kiarash Behrouzfar 1 2 , Kimberley Patrick 1 2 3 , Steve E. Mutsaers 3 , Grant Morahan 4 , Richard A. Lake 1 3 , Scott A. Fisher 1 2 3
  1. National Centre for Asbestos Related Diseases, Perth, Western Australia, Australia
  2. School of Biomedical Sciences, The University of Western Australia, Perth, Western Australia, Australia
  3. Institute for Respiratory Health, The University of Western Australia, Perth, Western Australia, Australia
  4. Centre for Diabetes Research, , Harry Perkins Institute of Medical Research, Perth, Western Australia, Australia

Background

Mesothelioma is a rare, but lethal, therapy resistant cancer with poor prognosis. Among current treatment options, combination of the immune checkpoint inhibitors ipilimumab and nivolumab have significantly improved treatment outcomes of mesothelioma. However, positive treatment responses remain restricted to a subset of patients, highlighting the need to identify biomarkers or gene signatures that better predict treatment response.  

Here, we exploited our well characterised, CC-MexTAg mesothelioma model to study the influence of tumour gene expression on asbestos induced mesothelioma development.

 

Methods

In this study, we performed comprehensive gene expression and immunohistochemical analyses on 167 tumour samples harvested from 105 mice representing 35 genetically distinct CC-MexTAg strains. We used Weighted Gene Co-Expression Network Analysis (WGCNA) to identify 20 key hub genes and utilised their human homologues to interrogate two large independent human mesothelioma datasets. Throughout this process, we conducted Univariate and LASSO cox regression analyses to develop the refined cox model containing the optimal set of genes to predict patient risk and prognosis. To test the gene signature as an independent prognostic predictor, we performed multivariate cox regression analysis while adjusting for sex, age and histological subtypes.

Results

Principal component analysis of RNA sequencing data revealed a distinct cluster of ‘inflamed’ tumours characterised by a high level of CD4 T and B cells. Using the key 20 hub genes associated with the inflamed cluster to interrogate human mesothelioma datasets allowed us to construct a 6-gene prognostic signature associated with poor survival outcome and high predictive performance (AUC=0.79 and AUC=0.71 at 2 years survival), which was successfully validated by TCGA and Bueno et.al human mesothelioma datasets. Moreover, multivariate cox regression analysis identified the gene signature as an independent predictor of patient survival after controlling for sex, age and histological subtypes in both datasets.

Conclusion

Our analysis revealed a 6-gene mesothelioma-specific prognostic signature that accurately predicted survival outcome in multiple independent, human mesothelioma datasets. Identification of high-risk patients prior to treatment has the potential to better stratify patient treatment selection and improve therapeutic outcomes.